Resource Management for QoS-Guaranteed Marine Data Feedback Based on Space–Air–Ground–Sea Network

More developed marine sensors for various applications has induced a rapid increase in marine data. The feedback from these marine data becomes challenging due to the backward marine communication techniques. The space–air–ground–sea integrated network (SAGSIN) provides a possible solution to solve this challenge by making use of the advantages of different networks. However, how to coordinate these networks and manage heterogeneous resources to satisfy the communication requirements of different marine applications remains to be solved. In this article, we investigate the resource management problem of SAGSIN for marine applications. A resource management architecture is proposed in which software-defined networking (SDN) controllers are employed. Based on this architecture, heterogeneous resources can be scheduled, and the data from devices with different communication modes can be transmitted via SAGSIN without changing the communication mode of the devices. We further propose two multiagent deep reinforcement learning resource management schemes to help individual devices find optimal access and resource allocation decisions to feed their data back to the terrestrial data centers. The design of these proposed schemes fully considers the scarce communication resources of marine scenarios, which makes data feedback more communication efficient while satisfying quality of service (QoS) requirements. Simulation results show that the improved MA_SDN_Centralized resource management scheme can significantly reduce the blocking probability of the system with guaranteed QoS, while reducing the communication overhead of learning.